# Predicting Invasive Non-mucinous Lung Adenocarcinoma IASLC Grading: A Nomogram Based on Dual-energy CT Imaging and Conventional Features

## TL;DR

This study creates a model using dual-energy CT scans and clinical data to predict the aggressiveness of a type of lung cancer called invasive non-mucinous adenocarcinoma.

## Contribution

A new nomogram model combining dual-energy CT imaging and clinical features is developed to predict IASLC grading of INMA preoperatively.

## Key findings

- Smoking history, lobulation sign, air bronchogram, and DECT parameters like eff-Z and IC are independent predictors of INMA IASLC grading.
- The nomogram model achieved an AUC of 0.804 with high specificity and moderate sensitivity.
- The model shows potential for non-invasive preoperative evaluation of INMA aggressiveness.

## Abstract

肺腺癌是非小细胞肺癌（non-small cell lung cancer, NSCLC）重要的病理组织学亚型。而肺浸润性非黏液腺癌（invasive non-mucinous pulmonary adenocarcinomas, INMA）因其显著异质性及组织学成分多样性，患者预后往往较差。建立INMA组织学分级系统是评价其恶性程度的关键。2021年国际肺癌研究协会（International Association for the Study of Lung Cancer, IASLC）提出新的组织学分级系统可以更好地对INMA患者进行预后分层。本研究旨在通过双能计算机断层扫描（dual-energy computed tomography, DECT）参数、分形维数（fractal dimension, FD）、临床特征及常规CT参数建立可视化列线图模型来术前预测INMA IASLC分级。

回顾性纳入2021年3月至2025年1月术前行DECT的INMA患者112例。根据IASLC分级将患者分为低-中级别组和高级别组。收集患者临床特征及常规CT参数，包括基线特征、生化标志物及血清肿瘤标志物。双能CT衍生参数，包括碘浓度（iodine concentration, IC）、有效原子序数（effective atomic number, eff-Z）和标准化碘浓度（normalized iodine concentration, NIC），采集并测定NIC比（NIC ratio, NICr）和FD。采用单因素分析比较两组在传统特征及双能CT衍生参数上的差异，将有统计学意义的变量纳入多因素Logistic回归分析，建立临床资料、常规CT参数及双能CT衍生参数的列线图模型并筛选INMA IASLC分级的独立预测因子；利用受试者工作特征（receiver operating characteristic, ROC）曲线分析评估判别能力。

多因素分析显示吸烟史[优势比（odds ratio, OR）=2.848, P=0.041]、分叶征（OR=2.163, P=0.004）、支气管充气征（OR=7.833, P=0.005）、动脉期eff-Z（OR=4.266, P<0.001）及动脉期IC（OR=1.290, P=0.012）是预测INMA IASLC分级的独立影响因素，基于上述指标构建的列线图模型预测性能最佳，曲线下面积（area under the curve, AUC）达0.804（95%CI: 0.725-0.883），特异度和灵敏度分别为85.3%和65.7%。

基于临床特征、影像学特征及能谱CT衍生参数的列线图模型在INMA IASLC分级的术前无创评估中具有较大的应用潜力。

Baseline clinical data of 112 patients with INMA

Comparison of imaging features of INMA in low-to-moderate and high grade groups

A, B: Effective atomic number (eff-Z is 8.09 and 8.16 respectively) of AP and VP; C, D: IC pseudo-color images of AP and VP (IC is 7.85 and 8.93 mg/cm3 respectively); E, F: Spectral curve images of AP and VP (K is 1.06 and 1.21 respectively); G: The fractal dimension analysis of the lung window setting revealed a fractal dimension of 1.67; H: Hematoxylin and eosin staining (×100).

A, B: Effective atomic number (eff-Z is 9.38 and 9.03 respectively) for AP and VP; C, D: IC pseudo-color images of AP and VP (IC is 31.63 and 24.82 mg/cm3 respectively); E, F: Spectral curve images of AP and VP (K is 3.62 and 2.98); G: The fractal dimension analysis of the lung window setting revealed a fractal dimension of 1.74; H: Hematoxylin and eosin staining (×100).

Comparison of spectral parameters of INMA in low-to-moderate and high grade groups

DECT related parameters of ICC

Independent risk factors for INMA IASLC grading

A: Predictive nomogram for INMA IASLC grading plotted with smoking history, lobulation sign, air bronchogram, eff-Z and IC of AP; B: ROC curves for each predictor assessing and comparing the diagnostic efficacy of nomogram models; C, D: Calibration and decision curves for the nomogram model predicting INMA IASLC grading. ROC: receiver operating characteristic.

Diagnostic efficacy of independent risk factors and combined models